摘要： 在马尔可夫随机场(MRF)和概率理论的基础上，提出局部区域能量最小化模型，将传统基于像素的分割转化为基于区域的分割，能减小均匀区域中的误分类率。在该模型和MRF模型下，使用ICM算法、Gibbs采样算法、Metropolis采样算法对图像进行分割，结果表明该模型能取得更精确的分割结果，可有效拟制图像噪音和纹理对分割的影响。

关键词: 图像分割, 能量最小化, 马尔可夫随机场, 最大期望算法

Abstract: This paper presents local region energy minimization model based on Markov Random Field(MRF) and probabilistic theory. The model converts traditional segmentation based on pixel into that based on region. It can reduce misclassification rate among the smooth area. The model is compared with the MRF model using ICM algorithm, Gibbs sampler algorithm and Metropolis sampler algorithm to segment image. Results show that the proposed energy model is able to obtain more accurate segmentation result and also can effectively restrain effect of image noise and texture for segmentation.

Key words: image segmentation, energy minimization, Markov Random Field(MRF), Expectation Maximization(EM) algorithm

中图分类号: 

• TP393